Direct address to camera

Audio says the exact line, transcribed word for word, with eye contact held to the end.

Audio says the exact line, transcribed word for word, with eye contact and lips closing on the p of slop.

Audio present, exact line, accurate phoneme lip-sync.

Audio present, mouth shapes match the stated line.

Audio present, mouth shapes match the stated line.

Exact line, phoneme-accurate sync.

Exact line, tight phoneme-level sync.

Speaks the requested line accurately with tightly synced mouth movement, then repeats a fragment of it ('you do not have to') a second time -- minor script deviation, everything else clean. Real, accurate speech audio confirmed.

Silent clip (no native audio); natural, well-coordinated speech articulation.

The lips follow the audio, but the line comes out as "We test the flop", heard that way by two separate transcriptions; the clip also cuts from a medium shot to a close-up at 1.25 seconds.

Brief gibberish audio before the intended line lands with accurate phoneme lip-sync.

Says "We test we the slop" before delivering the full line, transcribed word by word; the lips follow both attempts.

The woman never speaks or moves her mouth at all for the full 8s despite the prompt requiring a specific spoken line -- verified via frame extraction (multiple frames all show a fully closed, neutral mouth): confirmed, not audio-related (native_audio=0 explains silence, not the total absence of articulation).

Lips mouth the line correctly, but NO audio was generated at all. Fidelity docked for the missing spoken component; audio dimension left unscored (see model note), not failed.

FAILURE MODE: no audio track at all; mouth does generic open/round/close flapping unrelated to the intended line.